I have a problem with understanding how the resolution of the identity of an operator is presented in some literature for physicists.

I'm a student of mathematics, and I understand the notion of a spectral measure (which is somethimes called the resolution of identity) and also have some knowledge in spectral theory (for normal operators).

Here is my brief explanation what do I understand:

Let $H$ be Hilbert space (with an inner product linear w.r.t 2nd coordinate) with an orthonormal basis $(e_{n})$ and define a linear operator (diagonal operator) $$ A = \sum_{i \geq 1} \lambda_i \left|e_i\right>\left<e_i\right|,$$ where $(\lambda_i)$ is a sequence of complex numbers (the properties of this sequences determine the properties of $A$ such as boundedness, selfadjointness, compactness etc.), to make sense of $A$ we assume that the above series converges in SOT, I also used Dirac bra-ket notation. The associated spectral measure of $A$ is defined via $$ E(\Delta) = \sum_i \mathbf{1}_{\Delta}(\lambda_i) \left|e_i\right>\left<e_i\right|,$$ where $\Delta$ is an element of Borel sigma field over the spectrum of $A$, and we have that $$\left<x| Ay\right> = \int_{\sigma(A)} \lambda \left<x| E(d\lambda)y\right> \ \ (x,y \in H).$$

Very often physicists would use the following notation for $A$ which acts on an element $\psi \in H$ $$ A\left|\psi\right> = \sum_{i} \lambda_i \left|i\right>\left<i\right|\left|\psi\right>.$$

My problems with notation

I started reading some notes, books about quantum field theory, and often it is written that, the identity operator $I$ on some (separable) Hilbert space $H$, has the expansion, called the resolution of the identity $$ I= \int dq^{\prime} \left|q^{\prime}\right>\left<q^{\prime}\right|,$$ I don't know whether it matters here but $\{\left|q\right>\}$ is supposed to be a complete set of states.

Reference: http://eduardo.physics.illinois.edu/phys582/582-chapter5.pdf bottom of p. 129.

My question

Is the notion of the above resolution of the identity the same as an integral w.r.t a spectral measure ($I$ is a diagonal operator)? If yes, how should I understand the above notation. If no, what do they actually mean by this resolution of the identity and how do they define this integral. I noticed that in a lot of book concerning quantum mechanics there are many calculations, but not very many definitions and assumptions, which makes stuff hard to understand for a mathematician.


2 Answers 2


The answer is Yes. The interpretation of the notation is quite straight forward: $dq'|q'\rangle\langle q'| = E(dq')$. We need to presume that $E$ is the spectral measure of an operator $Q' = \int q' E(dq') = \int dq'\, q' |q'\rangle\langle q'|$. The only aspect that doesn't necessarily mesh well with your question, as written, is the fact that you've defined a spectral measure $E$ only for operators whose spectrum consists of eigenvalues, since $A|e_i\rangle = \lambda_i |e_i\rangle$. Spectral measures can be defined for operators with any kind of spectrum, including continuous.

  • $\begingroup$ Thanks, I know that you have spectral theorem for any normal operator defined on a Hilbert space, however the notation for the resolution of identity in the example which I gave about was for the identity operator, which is in fact diagonal. One quick question so it supposed to be $$I= \int d q^{\prime} q^{\prime} \left|q^{\prime} \right> \left<q^{\prime}\right|$$ instead of $$I= \int d q^{\prime} \left|q^{\prime} \right> \left<q^{\prime}\right|$$? $\endgroup$
    – Eric
    Feb 6, 2015 at 16:06
  • $\begingroup$ Your first formula gives $Q'$ and not $I$, with $I$ given correctly by your second formula. The prototypical example of $Q'$ is the position operator in quantum mechanics. It has a simple continuous spectrum (in 1-dimension, that is) and that is why its "eigenvectors" are convenient for writing down a resolution of the identity. Essentially, you are representing identity using functional calculus $I = f(Q) = \int dq' f(q') |q'\rangle\langle q'|$, where $f(x) \equiv 1$. $\endgroup$ Feb 6, 2015 at 19:32
  • $\begingroup$ A good rule of thumb in interpreting this notation is to make sure that your interpretation works when the Hilbert space is finite dimensional and all integrals are replaced by a sums. $\endgroup$ Feb 6, 2015 at 19:34
  • $\begingroup$ Great answer! Thank you. For the identity we even don't need the functional calculus, we know that $Q^{\prime}$ as a selfadjoint operator admits a unique spectral measure $E$, thus by using the properties of a spectral measure we got $I=E(\sigma(Q^{\prime}))= \int_{\sigma(Q^{\prime})} 1 E(d\lambda)$, which can be of course written in a different notation which involves bra and kets. $\endgroup$
    – Eric
    Feb 6, 2015 at 19:47

See $\S$ 4.4 of de Madrid's "The role of the rigged Hilbert space in quantum mechanics"


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.